If there is a HTTP API to create a RASA app?

Rasa Open Source
1

I want the output of my RASA NLU model to pass the sentiment of the message too with the output.

As far as I know I can create a custom component to do the sentiment analysis but It will be for training only.

How to pass the sentiment of the message sent by user along with the output i.e. confidence scores.

Thank You

2

hey @Kavita, as far as what i can understand form your query i think so you want the sentiment analysis result too along with the Rasa NLU predicted Intent & Entities i.e you o/p should be like :

{

“text”: “show me chinese restaurants”,

“intent”: “restaurant_search”,

“entities”: [ { “start”: 8,

  "end": 15,

  "value": "chinese",

  "entity": "cuisine",

  "extractor": "ner_crf",

  "confidence": 0.854,

  "processors": []

}],

sentimentalData :
{

mood : “happy”

}}

correct me if i am wrong, i searched for the above operation, so following is what i suggest : You can override the Interpreter class to predict the sentimental Data, https://rasa.com/docs/core/_modules/rasa_core/interpreter/#RasaNLUHttpInterpreter

Altough its not the correct way to do so because we are trying to make changes to the existing code but you can give it a try :sweat_smile:

3

@JiteshGaikwad If I change the code I cannot use it in another PC or deploy it as I will have to copy/edit that specific file every time I install RASA NLU. Right??

There is no other way to create an app (to do what I want) in RASA right (without changing the rasa nlu code)???

4

hi , can you share the custom component code, currently i m working in custom NER , i want to integrate in pipeline i have no idea how to do it, can you help me please

5

@JiteshGaikwad Btw can we create a RASA app using the http API?

6

hey @Kavita can you please elaborate what you mean by Rasa App😅

7

hi @jiteshGaikwad can you help me out in custom component , i going to connect BERT has NER in rasa nlu pipeline, i have no idea how to do it. how rasa nlu takes the input from the BERT NER.is there any examples.Has @kavita working on sentimental analysis pipeline i m working on custom NER.

8

hey @vigneshgig i dont know about BERT, can you please tell abt that😅

9

BERT is a Bidirectional Encoder Transformer Like RNN or Attention Neural Network this is one kind of the network,Google is working on this BERT project, It can be used for multiple NLP Task.Now it is the ranking 1 Network for NLP task.It is trained with billon of data.For good Entity extraction we have to use lookup table which is manual work,So i decided to use BERT Network. Now I experimenting with it.In rasa nlu we are using NER_CRF only for entity extraction.Instead of NER_CRF i want to use BERT How can i do it. BERT output is for sentence “show me restaurant in Chennai”, BERT will give the output as [show,‘o’ ],[me,‘o’],[‘restaurant’,‘o’],[‘in’,‘o’],[‘chennai’,‘LOC’] LOC - location enitity.I want any example regarding custom pipeline and custom component. Thanking you, one more thing we can also integrate BERT with NER_CRF.So that can i change the NER_CRF file in rasa_nlu_master.

1 Like